ArkHive · open-source MCP servers

The memory your AI can't fake.

Where context gives meaning.

A tamper-evident, hash-chained ledger giving your agent persistent identity, auditable history, and cryptographic proof of everything it's done. Zero config, local-first, Apache-2.0.

Every entry is a hash-linked block, so nothing — not a hallucination, not a bad actor, not even you — can quietly rewrite what your agent knew or did, and one call proves the whole chain is intact. Give it a name and a covenant and your agent becomes someone that persists across sessions instead of a stateless goldfish. Before it acts, govern checks the decision against your constraints, so "oops, it refunded everyone" never ships. And every recall pulls what it already knows instead of re-deriving it — fewer tokens, faster turns. It all runs local: your data never leaves the machine, no API keys, no vendor lock-in. The accountability layer your agents were missing.

Get it on GitHub → Install ArkHive Algernon

Apache-2.0local-firstzero config

ArkHive MCP

Governed, tamper-evident memory for your agent. Five tools: birth · remember · recall · verify · govern.

birth — earn an identity: register a name + covenant, get an immutable soul.

remember — append a tamper-evident memory: who did what, with what data.

recall — fetch what the agent already knows instead of re-deriving it.

verify — prove the entire chain is unbroken, on demand.

govern — check a decision against your constraints before it acts.

Try it in 10 seconds — hosted, no key

# Claude Code:
claude mcp add --transport http arkhive https://arkhive.dondatabrain.com/mcp

# Any MCP client — HTTP endpoint:
https://arkhive.dondatabrain.com/mcp

Run it locally — your data never leaves the box

pip install arkhive-mcp
# just renamed — if pip can't find it yet, the old name still works:
# pip install humane-intelligence

Then add it to Claude Desktop (claude_desktop_config.json) or any MCP client:

{
  "mcpServers": {
    "arkhive": { "command": "arkhive-mcp" }
  }
}

Prefer a plain Python library?

pip install arkhive-ledger   # one file, zero dependencies
python -c "import arkhive; me = arkhive.birth('scout', ['do no harm']); print(arkhive.verify())"

Algernon MCP

Provider-agnostic fleet orchestration: split a big goal into tightly-scoped parallel sub-tasks, run them on cheap workers with your own LLM key, and stay free to think. Orchestrate a fleet. Keep your mind.

algernon_plan — decompose a goal into k tightly-scoped, independent sub-task prompts.

algernon_dispatch — run N tasks concurrently on a fleet of cheap workers and collect every result.

algernon_orchestrate — one shot: plan, fan out, and return the plan plus all results.

Install

# From source (today):
pip install git+https://github.com/sammyboi81/algernon

# PyPI (publishing this week):
pip install algernon-mcp

Source: github.com/sammyboi81/algernon

Want it done for you?

Both servers are free forever — self-host, no telemetry. When you'd rather not DIY:

Custom AI agent — $700 flat — a working MCP agent built for your workflow, installed into your Claude or ChatGPT in one call, done-for-you by the founder. Claim a slot →

ArkHive Enterprise — from $2,500 — hand-delivered install + pilot on your own server, with the team that built it. Contact us →

Built by the team behind InboxAxe — the governed AI marketing platform where nothing sends without your yes.

What stays private

These two servers are fully open source and self-hostable. The production DonDataBrain substrate they were grown in — the 2.29M-block flagship chain and the Senthar runtime — is licensed separately. Safety shouldn't be behind a paywall; the safety layer is the part we give away.